Thesis

Multi-view image synthesis techniques for 3D vision and free-viewpoint applications

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2012
Thesis identifier
  • T13266
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • The depth information in a scene for a stereo image is used in many image analysis and 3D video processing applications. Novel view synthesis draws a significant research interest because it can drive future 3DTV and free viewpoint video applications, which allows a viewer to perceive 3D depth scenery without wearing any special glasses. The main objective of the multi-view video system is to create another dimension to the viewer and provide 3D information such as depth. This thesis describes new approaches and methods for stereo matching and inter-view synthesis algorithms with application for 3D and free-viewpoint. A Depth Image Layers Separation (DILS) algorithm is proposed to efficiently synthesize the inter-view image based on layered disparity depth map representation through stereo matching and inter-view interpolation. The main idea of this approach is to separate the depth map into several layers of depth based on the disparity distance of the corresponding points. This technique is used to synthesize novel inter-view images based on disparity depth map layers representation. Simulation results show that the concept of depth layers separation is able to create inter-view images and can be integrated with other technique such as the disparity depth refinement and occlusion handlings processes. The DILS algorithm can be performed from a simple to sophisticated stereo matching techniques to synthesize the inter-view images. This technique leads to the second novelty method, Depth Layer Refinement (DLR) that uses the disparity depth layers to refine the disparity map. The main aim of this algorithm is to improve the raw disparity maps in the disparity refinement stage with a basic similarity metric of SAD in the stereo matching algorithm. The edge boundaries and discontinuities region are significantly improved with the proposed techniques compared to the state-of-the-art stereo matching algorithms. The third novelty proposed in the multi-view camera applications known as Multi-Level View Synthesis (MLVS). In this technique, the multi-view synthesis created based on a limited number of cameras to create dense images. The new structures and design are shown to offer improved performance and provide additional views with fewer cameras arrangement compared to the conventional high volume camera configurations for freeviewpoint video acquisition.
Resource Type
DOI
Date Created
  • 2012
Former identifier
  • 948340

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T13266 2021-07-02 Public Download